It’s no secret that when we interact with AI models like ChatGPT, Claude, or Gemini, it consume resources . This is because these resources are required for the AI to make its calculations, or to “think,” if you’d prefer. But how many resources does it consume? According to Google , a typical Gemini AI text prompt costs about five drops of water .
Google says AI prompts don’t consume much resources
Based on Google’s testing, “We estimate the median Gemini Apps text prompt uses 0.24 watt-hours (Wh) of energy, emits 0.03 grams of carbon dioxide equivalent (gCO2e), and consumes 0.26 milliliters (or about five drops) of water — figures that are substantially lower than many public estimates. The per-prompt energy impact is equivalent to watching TV for less than nine seconds.”
It sounds amazing. After all, the amount of hardware like GPUs needed to build and train an AI model are pretty outrageous. Plus, there are all these talks about how much energy it takes to run an AI data center. So, if Google says it only takes about five drops of water for a typical AI prompt, that’s quite a relief, right?
Unfortunately, experts disagree.
Experts are saying otherwise
Speaking to The Verge, experts are saying that Google’s claims are misleading . They also say that Google had left out key data points in its study . This includes omitting indirect water use in its estimates, for instance, the water used in the cooling systems that help keep servers from overheating.
Then, it seems that Google has also left out an important metric when it comes to power consumption and pollution. Right now, the study only shows a “market-based” measure of carbon emissions. However, this is based on commitments a company makes to support renewable energy growth on power grids.
The experts say that a better approach would be to use location-based measurements of carbon emissions. This is because certain locations might use different forms of energy generation. Using a one-size-fits-all approach would not necessarily be fair.
According to the experts, Shaolei Ren, an associate professor of electrical and computer engineering at the University of California, Riverside, and Alex de Vries-Gao, founder of the website Digiconomist and a PhD candidate at Vrije Universiteit Amsterdam Institute for Environmental Studies, “If you look at the total numbers that Google is posting, it’s actually really bad. This is not telling the complete story.”
